RESEARCH ARTICLE

Phenotype-driven chemical screening in zebrafish for compoundsthat inhibit collective cell migration identifies multiple pathwayspotentially involved in metastatic invasionViviana E. Gallardo1, Gaurav K. Varshney1, Minnkyong Lee2, Sujata Bupp2, Lisha Xu1, Paul Shinn3,Nigel P. Crawford2, James Inglese2,3 and Shawn M. Burgess1,*

ABSTRACTIn the last decade, high-throughput chemical screeninghasbecome thedominant approach for discovering novel compounds with therapeuticproperties. Automated screening using in vitro or cultured cell assayshave yielded thousands of candidate drugs for a variety of biologicaltargets, but these approaches have not resulted in an increase indrug discovery despite major increases in expenditures. In contrast,phenotype-driven screens have shown a much stronger success rate,which is why we developed an in vivo assay using transgenic zebrafishwith aGFP-markedmigrating posterior lateral line primordium (PLLp) toidentify compounds that influence collective cell migration. We thenconducted a high-throughput screen using a compound library of 2160annotated bioactive synthetic compounds and 800 natural products toidentify molecules that block normal PLLp migration. We identified 165compounds that interfere with primordium migration without overttoxicity in vivo. Selected compounds were confirmed in their migration-blocking activity by using additional assays for cell migration. We thenproved the screen to be successful in identifying anti-metastaticcompounds active in vivo by performing orthotopic tumor implantationassays in mice. We demonstrated that the Src inhibitor SU6656,identified in our screen, can be used to suppress the metastaticcapacity of a highly aggressive mammary tumor cell line. Finally, weused CRISPR/Cas9-targeted mutagenesis in zebrafish to geneticallyvalidate predicted targets of compounds. This approach demonstratesthat the migrating PLLp in zebrafish can be used for large-scale, high-throughput screening for compounds that inhibit collective cellmigrationand, potentially, anti-metastatic compounds.

KEY WORDS: Drug screening, Metastasis, Orthotopic implantation,Src, Tks5, Zebrafish

INTRODUCTIONCancer primarily kills bymetastasis,with at least 90%of cancer deathsbeing caused not by the primary tumor but by cancer invasion inremote tissues. Yet, metastasis remains the most poorly understoodcomponent of cancer pathogenesis because this process is inherently

difficult to model and study in vitro. The transition of a cancer from a‘localized’ tumor to a metastatic form requires the acquisition ofseveral cellular transformations, including migratory propensity andinvasiveness (Chaffer and Weinberg, 2011; Valastyan andWeinberg,2011), and these transformations are strongly influenced by thesurrounding normal, heterogeneous tissues.

High-throughput screens using in vitro or cell-based models thattypically target specific candidate genetic pathways have beendeveloped to identify drugs that can inhibit collective cell migrationin cancer metastasis (Chua et al., 2012; Quintavalle et al., 2011).These studies generated many ‘hits’; however, recent analysis hasdemonstrated that target-based screening has a very poor success ratewhen it comes to identifying potential therapeutic drugs (Swinney andAnthony, 2011). In contrast, phenotype-driven screening has a muchhigher rate of success (Swinney and Anthony, 2011); therefore, thecloser one can model the ‘natural’ environment of cell migrationin vivo, the more likely it is one will discover novel compounds withpotential therapeutic value.

Tissue opacity in most animal model systems makes real-timestudies of collective cell migration in morphogenesis and cancerdifficult. However, using transgenics and time-lapse imaging,the migration of the zebrafish posterior lateral line primordium(PLLp) has recently emerged as a powerful model to investigatethe molecular mechanisms and regulation of collective cellmigration (Aman and Piotrowski, 2010; Friedl and Gilmour,2009). The PLLp is a group of migrating cells that movesdorsally along the body of the fish, depositing clusters ofcells along the way that become the lateral line neuromasts(Aman and Piotrowski, 2011; Ghysen and Dambly-Chaudiere,2007). Parallels can been drawn between the collectivemigration of the PLLp cells and the behavior of invasivecancer cells. In both events, cells delaminate, acquire migratorybehavior, and progress through extracellular matrix to reach adistant target.

Our previous studies showed that a large number of genesexpressed specifically in the migrating primordium have commonroles in collective cell migration and cancer progression (Gallardoet al., 2010). In addition, others have shown that the main signalingpathways responsible for the regulation of cell mobilization areactive during both development and tumor metastasis (Yang andWeinberg, 2008). Therefore, it is possible that compounds thatinhibit the natural progression of the posterior lateral line (PLL) alsohave potent anti-metastatic activity.

Given the cellular and molecular parallels between lateral linedevelopment and metastasis, we developed a whole-organism-basedchemical screening strategy combined with automated fluorescencemicroscopy to identify small-molecule modulators of zebrafish lateralline migration. We took advantage of the transgenic reporter lineReceived 7 October 2014; Accepted 19 March 2015

1Developmental Genomics Section, Genome Technology Branch National HumanGenome Research Institute, National Institutes of Health, Bethesda, MD 20892,USA. 2Genetics andMolecular Biology Branch, National HumanGenomeResearchInstitute, National Institutes of Health, Bethesda, MD 20892, USA. 3Department ofPre-Clinical Innovation, National Center for Advancing Translational Sciences,National Institutes of Health, Rockville, MD 20850, USA.

*Author for correspondence (burgess@mail.nih.gov)

This is an Open Access article distributed under the terms of the Creative Commons AttributionLicense (http://creativecommons.org/licenses/by/3.0), which permits unrestricted use,distribution and reproduction in any medium provided that the original work is properly attributed.

565

© 2015. Published by The Company of Biologists Ltd | Disease Models & Mechanisms (2015) 8, 565-576 doi:10.1242/dmm.018689

Disea

seModels&Mechan

isms

cldnb:EGFP (Haas and Gilmour, 2006), which expresses GFP in allof the cells of the PLL and, using this reporter line, screened acollection of drugs and other bioactive compounds (Sigma LOPAC1280), a collection of 800 natural products (NatProd Collection), andthe GSK Published Kinase Inhibitor Set (PKIS) to identifycompounds that inhibited collective cell migration. We identified165 compounds that interfered with primordium migration withoutovert toxicity in vivo. We identified several kinase inhibitors,flavonoid-derivatives and antioxidants that all disrupted primordiummigration. We also confirmed, in a mouse tumor model, that the Srcinhibitor SU6656 also specifically decreases tumor metastasis.Finally, we showed that target pathways can be quickly geneticallyvalidated in zebrafish by using CRISPR/Cas9 targeting. Takentogether, our approach suggests that the migrating PLLp in zebrafishcan be used for large-scale, high-throughput screening for inhibitorsof collective cell migration.

RESULTSScreening for cell-migration inhibitorsWe developed a whole-organism-based chemical screening strategyto rapidly identify novel small-molecule modulators of cell migrationduring zebrafish PLL formation. We used cldnb:EGFP embryos toscreen the LOPAC 1280 library, the PKIS and the NatProd collectionfor compounds that alter the migration of the lateral line.

At 20 h post-fertilization (hpf) (which coincides with the onset ofthe primordium migration), cldnb:EGFP embryos were manuallyarrayed into 96-well dishes (two embryos per well) using a 200-µlwide-bore pipette tip and treated with test compounds at a finalconcentration of 10 μM. All plates contained five negative controlwells (1% DMSO) and five positive control wells (K252a, the broadactivity kinase inhibitor previously determined to arrest PLLpmigration) (Fig. 1), and migration for each compound was scored incomparison to the control wells.

Embryos were monitored over a period of 2 days for correctlateral line migration. A fluorescence image of each well wascaptured by using an automated imaging system at 16 and 28 h post-incubation with the test compounds (Fig. 1D). In addition tomigration disruption, drug-induced lethality and other visiblephenotypes were scored. We identified all PLLp cells that had notmigrated to the tail of the fish by 48 hpf. Compounds that scoredpositive for inhibiting migration were retested and all compoundsthat scored positively twice were considered as potentially anti-metastatic compounds.

The majority (1543 compounds, 74.18%) of the 2080 compoundstested from the LOPAC and NatProd libraries did not causean observed phenotype in 2-dpf cldnb:EGFP embryos. Of theremaining compounds, 21% were toxic at 10 µM. We retested thetoxic compounds at final concentrations of 1 and 5 μM. On the otherhand, more than 90% of compounds initially tested with thePKIS were toxic at 10 µM. Consequently, we screened this libraryat 0.5, 1 and 5 µM.

Altogether we identified 165 compounds that interfered withprimordium migration without causing overt toxicity (gross develop-mental defects) in the larvae. The screen resulted in a compound list ofactivity annotations related mainly to phosphorylation, neuro-transmission and cell signaling. Overall, 5.57% of compoundsscreened disrupted collective migration of the primordium and,therefore, the development of the PLL. Fig. 2 shows representativePLL phenotypes observed in the primary screen. Exposure to a varietyof inhibitor classes resulted in disruption ofmigration. For example, thePPAR-alpha receptor agonist Fenofibrate (Fig. 2B) and the cyclin-dependent kinase (CDK) inhibitor Kenpaullone (Fig. 2F) induce aphenotype with the premature deposition of a terminal neuromasts,whereas treatments with the cell-cycle inhibitor methylselenocysteine(Fig. 2C), theDNA-topoisomerase-II inhibitorEllipticine (Fig. 2D) andthe PARP inhibitor 6(5H)-Phenanthridinone (Fig. 2E), all disruptedlateral line migration compared with the DMSO controls (Fig. 2A).Treatment with the antioxidant Purpurogallin-4-carboxylic acid(Fig. 2G) or the flavonoid-derivative, 4′-Methoxyflavone (Fig. 2H)completely abolished the PLL. Quantitative analysis of the lastdeposited neuromast at 48 hpf is shown in Fig. 2I. Average positionof deposited neuromast was estimated relative to the distance from theotic vesicle (ov) to the tip of the tail as performed in Matsuda et al.(2013).

Our screening identified a wide variety of pathways whoseinhibition had potentially anti-metastatic effects (supplementarymaterial Table S1). Additionally, the following six broadcategories were targeted by two or more compounds: cellsignaling, gene regulation, ion channels, lipid metabolism,

TRANSLATIONAL IMPACT

Clinical issueCancer is a leading cause of death worldwide. As high as 90% ofcancer deaths are a result of metastasis, yet this remains the mostpoorly understood component of cancer pathogenesis. The currentpreclinical pipeline for target-driven drug discovery involves multiplerounds of in vitro biochemical and cell-based assays followed byin vivo studies in animal models, and finally trials in humans. Thisprocess typically takes 12-15 years before drugs reach the market andis expensive, limiting the number of compounds that can effectively betranslated into therapeutic use. Over the past decade, the focus ofdrug screens has been on high-throughput screens using in vitroassays or cell-based models that target specific candidate pathways,with the aim of inhibiting cancer metastasis. These studies havegenerated thousands of candidate drugs for a variety of biologicaltargets; however, these approaches have had very poor success rateswhen it came to therapeutic drugs because they generally lackedrelevant whole-organism physiology. Most of the positive in vitroresults were not replicated when tested in vivo. Thus, an in vivophenotype-driven screen in a whole-animal model should providebetter targets for therapeutic intervention with a much strongersuccess rate, shortening years of research and increasing cost-effectiveness.

ResultsIn this study, the authors developed a robust in vivo assay usingtransgenic zebrafish to mark the migrating posterior lateral lineprimordium as readout for inhibition of collective cell migration. Via ahigh-throughput screening protocol, the authors identified a number ofcompounds, which included novel flavonoid-derivative molecules and acluster of structurally related kinase inhibitors that interfered withprimordium migration without overt toxicity in vivo. The goal ofidentifying cell-migration inhibitors is to find compounds that have anti-metastatic activity. Demonstrating the utility of this approach, the authorsconfirmed, by performing orthotopic tumor implantation assays in mice,that inhibition of the Src pathway decreases tumor metastasis in vivo.Finally, the authors used CRISPR/Cas9 targeted mutagenesis inzebrafish to validate targets of essential genes involved in cellmigration, showing that zebrafish can be used to rapidly confirm themolecular targets of inhibitory compounds.

Implications and future directionsThis study highlights the utility of the zebrafish migrating primordium asan in vivo large-scale, high-throughput screening system for cell-migration inhibitors. This study also demonstrates that this screen canbe used to successfully identify both compounds and new pathways fortargeting cancer metastasis. In addition, this approach represents astarting point for future in-depth studies to develop new therapeuticstrategies for cancer.

566

RESEARCH ARTICLE Disease Models & Mechanisms (2015) 8, 565-576 doi:10.1242/dmm.018689

Disea

seModels&Mechan

isms

neurotransmission and natural products of undefined activity.There were compounds with interesting phenotypes notspecifically related to migration; for example, the two EGFRinhibitors Tyrphostin AG1478 and GW2974 generatedsupernumerary neuromast numbers unrelated to a migrationphenotype (supplementary material Fig. S1), supporting the ideathat ErbB signaling is required in glial cells to repress theprecocious neuromast formation from cells that lay betweenneuromasts (Grant et al., 2005).

Small-molecule compounds affect collective migration ofthe PLLpTo confirm the result of the primary screen we tested a group of 18compounds, selecting the kinase-related molecules and the mostactive natural products. We grouped these compounds by thegenes that they were designed to target (supplementary materialTable S2). This kinase-compound list contained three Src-familykinase inhibitors: Emodin, SU6656 and 7-cyclopentyl-5-(4-phenoxy)phenyl-7H-pyrrolo[2,3-d]pyrimidin-4-ylamine (RBI);two PDGFR inhibitors, Tyrphostin A9 and SU 4312; and the CDKinhibitors Kenpaullone and Indirubin-3′-oxime, among others.Five highly active compounds in the natural products collectionwere

identified in the screen.The list contained three flavonoid-derivatives –genistein, 7,3′-dimethoxyflavone (previously identified as an anti-invasive compoundbyParmaret al., 1994) and4′-Methoxyflavone; theantioxidant Purpurogallin-4-carboxylic acid and the aldose-reductase

inhibitor 4-Naphthalimidobutyric acid – all of which strongly inhibitedthe migration of the PLL.

Identified compounds inhibit cell migration in other contextsTo provide independent verification that we are identifyingcompounds that inhibit cell migration, we performed an in vivoneutrophil migration assay in zebrafish, in which specific non-invasive damage to the lateral line neuromasts by using Cu2+ (inthe form of CuSO4) induces an acute inflammatory response(Fig. 3A) (d’Alencon et al., 2010). Compounds that inhibit cellmigration would prevent the invasion of neutrophils into thedamaged lateral line neuromasts. We used the Tg(mpx:GFP)transgenic fish, which allowed us to track tissue damage by visualobservation of GFP expressed in the neutrophils. We exposed72 hpf Tg(mpx:GFP) larvae to the 15 most effective (based onarrest strength and least toxicity) inhibitors of the LOPAC1280library (supplementary material Fig. S2) and four compounds ofthe NatProd library (supplementary material Table S3) for 30 minto allow the penetration of the drugs into larval tissues, followed bythe addition of Cu2+ for 40 min. As expected, the DMSO controlsshowed the normal distribution of neutrophils primarily localizedin the ventral trunk and tail. However, in Cu2+-treated larvae,neutrophils clustered around the neuromasts as part of theinflammatory response. In drug-treated larvae, all of the selectedcompounds tested, including the positive controls of K252a,exhibited statistically significant inhibition – on the basis of

Fig. 1. Overview of the in-vivo drugscreening strategy in zebrafish. (A) TheLOPAC1280, NatProd and PKIS librarieswere screened for cell-migration inhibitorsusing 20 hpf cldnb:EGFP embryos that wereGFP-positive for the migrating PLLp (markedwith small arrow). (B) Compounds weretransferred by ultrasonic dispenser from 384-well plates to 96-well plates so that, when200 µl of embryo medium was added, thecompounds were at a final concentration of10 μM (NCATS, NIH). (C) Each platecontained five negative (1% DMSO) and fivepositive (0.01-1 μM K252a) control wells. Twoembryos were placed manually into each wellof the 96-well plates and PLL developmentwas followed through 48 hpf. (D) Automatedimage capture of zebrafish embryos wasperformed using an iCys® Research ImagingCytometer and images were scored forcompletion of PLLp migration.

567

RESEARCH ARTICLE Disease Models & Mechanisms (2015) 8, 565-576 doi:10.1242/dmm.018689

Disea

seModels&Mechan

isms

neutrophil cell counts – within a defined region surroundingthe neuromast (supplementary material Fig. S2). In contrast,larvae pre-treated with hydrocortisone (a compound that scorednegative in the PLLp assay) did not exhibit significant inhibition inchemically induced inflammation (ChIn) assays (data notshown), as previously described in d’Alencon et al., 2010.Quantitative analysis for a subset of tested drugs is shown inFig. 3B.

Comparison of compounds identified by in vivo screeningwith traditional in vitro cell-based metastasis modelsTo further explore the generality of our results, we tested the effect ofthe selected compounds in in vitro cell-migration or cell-invasionassays by using melanoma-derived cell lines. We tested human

501-melmelanomacellswith kinase inhibitors from the LOPAC1280library for further evaluation. Eleven of the compounds causedsignificant inhibition of cell migration. The remaining compoundsdid not show significant inhibition compared to the DMSO control.Assays were repeated three times and representative pictures areshown in Fig. 3C. Quantification of the cell migration assay is shownin Fig. 3D. For invasion assays, we used hepatocyte growth factor(HGF) to induce migration, as it has been shown to stimulate theinvasive potential of melanoma cells (McGill et al., 2006). ForMatrigel invasion assays, we observed that treatment of 501-mel cellswith the tested drugs resulted in inhibition of invasion of these cellsinduced by HGF. Quantification of four compounds, K252a (a broadkinase inhibitor), Wartmannin (PI3K), Emodin (Src) and SU6566(Src), is shown in Fig. 3E.

Fig. 2. Examples of PLL migration phenotypes observed in the screen. Phenotypes of 48 hpf zebrafish embryos treated with small molecules that had beenidentified in the screen. (A)DMSOcontrol. The redbox indicates thesuccessfulmigrationof the terminal neuromasts to theendof the tail.Cldnb:EGFPembryos treatedwith 10 μMofSe-(methyl)selenocysteine (C), Ellipticine (D) or 6(5H)-Phenanthridinone (E) showa delay in primordiummigration and, therefore, have disrupted lateralline formation compared with theDMSOcontrol (A,I), whereas treatment with Fenofibrate (B) or Kenpaullone (F) causes an increase in the neuromast deposition rate,resulting in a phenotypewith premature deposition of terminal neuromasts. Treatmentswith the flavonoid-derivativemoleculesPurpurogallin-4-carboxylic acid (G) and4′-Methoxyflavone (H), almost completely abolished the posterior lateral line formation, generating PLL phenotypes of 1 (G) and 3 (H) deposited neuromasts,respectively. Arrows indicate the position of the PLLp relative to the fully successful migration marked by the red box in A; asterisks indicate the position of the lastdepositedneuromast inGandH. (I) Quantification of the averageposition of the final deposited neuromast in drug-treatedembryos relative to the distance from the oticvesicle (ov) to the tip of the tail. Experiments were carried out using 15 larvae per condition. Error bars indicate +s.d. Compounds showed a variety of effects, frominhibiting both neuromast numbers and migratory behavior (e.g. Purpurogallin-4-carboxylic acid) to only slowing migration (Fenofibrate).

568

RESEARCH ARTICLE Disease Models & Mechanisms (2015) 8, 565-576 doi:10.1242/dmm.018689

Disea

seModels&Mechan

isms

Fig. 3. Confirmation of active compounds. (A) Schematic view of a 3 day post-fertilization (dpf ) larva. The boxed area corresponds to the horizontalmyoseptum (dotted red lines) and shows the area where leukocytes were counted in all quantification experiments. (B) Quantification of infiltratingleukocytes in the lateral line after treatments with diverse inhibitors. The assays were carried out by manually counting leukocytes recruited to the lateralline neuromasts after Cu2+ treatment (10 μM) as described in Results. In this experiment, drugs were added 30 min prior to Cu2+ treatment. The graphshows average leukocyte numbers in the lateral line in negative controls (DMSO), in Cu2+ (as CuSO4) and in Cu2+-plus drug-treated fish. Leukocytequantification shows significant reduction (***P<0.001) in neutrophil recruitment to damaged PLL neuromasts in larvae exposed to the compounds testedcompared with Cu2+-exposed larvae, suggesting an inhibition of the neutrophil migration. Experiments were carried out with 15 larvae per condition.(C) Effects of kinase inhibitor compounds from LOPAC1280 on in vitro melanoma cell migration/invasion. 501-mel human melanoma cells were treated withDMSO (negative control) or the c-Met inhibitor K252a (positive control), and with the drugs indicated for 16 h and then assayed for motility asdescribed in Materials and Methods. Representative data from independent experiments are shown. Only compounds that had a negative effect onmigration are shown. (D) Quantification of 501-mel cell migration assays. The migratory cells were counted and the results expressed as the mean numberof migratory cells from two independent experiments. Measurements were calculated relative to control (DMSO). (E) Comparative effects of tested drugs onthe cell invasion capacity of melanoma cells. 501-mel cells were treated with DMSO, K252a, Wortmannin, Emodin and SU6656 and then plated on Matrigelinserts in presence of HGF at 50 ng/ml for 24 h. The invasive cells were counted and the results expressed as the mean number of migratory cells infive random microscopic fields. Measurements were calculated relative to control (DMSO). Data collected were from two independent experiments.***P<0.001; *0.01<P<0.05.

569

RESEARCH ARTICLE Disease Models & Mechanisms (2015) 8, 565-576 doi:10.1242/dmm.018689

Disea

seModels&Mechan

isms

Using Trypan Blue, we verified that treatment of 501-mel cellshad no significant effect on cell viability for any of the tested drugs(data not shown). Together, these results confirmed many of thecandidate drugs from our screen work when exposed to traditionalcell-based cancer invasion models. More importantly, we were ableto identify many compounds that worked in vivo, but would nothave been identified in a cell-based screen (supplementary materialTable S3).

Using CRISPR/Cas9 in zebrafish embryos to rapidly validateinhibitor targetsThe Src-Tks5 pathway has been demonstrated to be involved in theregulation of the migration of neural crest cells in zebrafish (Murphyet al., 2011), in macrophage invasive behavior (Burger et al., 2011)and in cancer cell invasion (Blouw et al., 2008; Courtneidge, 2012).Not surprisingly, we identified in our screening three Src inhibitors(supplementary material Table S1) that interfere with PLLpmigration. When tested compounds were added either at 20 hpf(when primordium migration begins; data not shown) or 36 hpf(when primordium migrated beyond the yolk extension and 2-3neuromasts have already been deposited), we observed in all cases adisruption in the migration of the primordium compared to theDMSO control (Fig. 4). As a result, Emodin-, RBI- and SU6656-treated embryos lacked the most caudal PLL neuromasts. Typically,embryos had only the first five or six neuromasts, with closerdeposition of the neuromasts (Fig. 4E). Although the Src pathwayhas been widely implicated in diverse cellular processes, Src activityhas neither been shown to be involved in PLLp migration duringdevelopment nor have Src inhibitors been shown to disrupt thisprocess. Since the three Src inhibitors had been independentlyidentified in the screen, evidence is strong that Src is the true target;however, one major consideration when screening inhibitors is the

significant issue of ‘off-target’ effects for many of the compounds.Morpholinos have a known artifact that impacts lateral linemigration, precluding their use, but CRISPR/Cas9 has beenshown to be a very effective technique for targeting genes inzebrafish (Hwang et al., 2013; Jao et al., 2013). To confirm that itwas truly inhibition of the Src pathway that impacted lateral linemigration, we used CRISPR/Cas9 to target src or a downstreamsubstrate of src, tks5a. When single-guide RNAs (sgRNAs)targeting src and tks5a (but not unrelated control sgRNAs) wereinjected into the cldnb:EGFP transgenic line, lateral line migrationwas blocked in a manner similar to that of Src inhibitors ormorpholinos targeting tks5a (data not shown) (Murphy et al., 2011).At 48 hpf, embryos injected with CRISPR sgRNAs targeting srcshowed a gross disruption of primordium migration (Fig. 5B). Src-mutant embryos showed strong migratory inhibition and depositionof very few – typically one or two – neuromasts. We observed asimilar but less-severe phenotype in tks5a-CRISPR-injectedembryos (F0) and tks5a−/− F1 homozygous mutants (Fig. 5C,D).At 48 hpf, tks5a−/− embryos showed delayed migration of theprimordium and PLL morphogenesis. Other defects were alsovisible in mutants, such as: smaller heads with small eyes, edemaand overt delay in appearance of pigment cells in the tail (Murphyet al., 2011). Compound heterozygous tks5a mutants containedsmall indels in exon 8 and exon 12. Exon 8 carried an 8-bp insertionand 1-bp deletion, whereas exon 12 carried a 21-bp insertion and 4-bp deletion (Fig. 5E). The more-drastic phenotype observed in src-CRISPR-injected embryos compared with tks5a−/− mutants couldbe because Src is located upstream of Tks5, regulating otherbiological pathways beyond those directly utilizing the tks5a gene.This demonstrates how quickly candidate targets can be confirmed(or identified in the case of multiple known targets of an inhibitor) inzebrafish by using CRISPR/Cas9.

Fig. 4. Blocking the Src signaling pathway inhibits collective cell migration in vivo. Cldnb:EGFP embryos were treated at 36 hpf (primordium has migratedbeyond the yolk extension, dashed line) with the c-Src inhibitors Emodin (B), RBI (C) or SU6656 (D). Treated embryos show a disruption of the primordiummigratory ability compared with DMSO control (A), resulting in a phenotype with a premature deposition of a terminal neuromasts (L6) in all conditions (B-D).(E) Quantification of average position of the last five deposited neuromasts in drug treated-embryos relative to the distance from the ov to the tip of the tail.Arrows indicate the position of the PLLp. Experiments were carried out with ten larvae per condition. Error bars indicate +s.d.; ***P<0.001.

570

RESEARCH ARTICLE Disease Models & Mechanisms (2015) 8, 565-576 doi:10.1242/dmm.018689

Disea

seModels&Mechan

isms

In addition, to avoid the off-target effects of Src inhibitors andvalidate the role of Src on PLLp migration, we also carried out rescueexperiments (Fig. 6). Overexpression of src mRNA was able topartially rescue primordium migration defects induced by RBI(Fig. 6C,D) or SU6656 (Fig. 6E,F). Furthermore, overexpression ofsrc reduces developmental abnormalities induced by thesecompounds. However, we did not see a significant rescue ofphenotype in injected embryos treated with Emodin compared withWT (Fig. 6A,B). The possible explanation for this observation is thatEmodin could be inhibiting other signaling pathways in addition to theSrc familykinases involved in thePLLdevelopment preventing rescue(Shrimali et al., 2013; Wei et al., 2013). As expected, overexpressionof src following DMSO treatment showed a normal lateral linephenotype (Fig. 6G) and was not able to rescue the LL phenotypeinduced by other signaling pathway inhibitors (data not shown).

Inhibition of Src activity through SU6656 decreases tumormetastasis in vivoKey to the value of performing an in vivo screen in zebrafish forcompounds that inhibit cell migration is to prove that identifiedcompounds can inhibit cancer invasion in vivo. To demonstrate thatinhibitingmigration of the zebrafish PLLp can be used to identify anti-metastatic compounds we selected the Src inhibitor SU6656, whichhas a published effective pharmacologic dose in mice (Rehni andSingh, 2011; Rehni et al., 2012), for follow-up studies in a mouse

tumor model. First, we tested the effect of SU6656 on the migrationand proliferation of human 501-mel melanoma and mouse mammarytumor 4T1 cells in transwell assays. SU6656 impaired the migrationof both cell lines in a dose-dependent manner, with EC50 values of8.2 μM and 7.2 μM, respectively (supplementary material Fig. S3A).Likewise, SU6656 treatment reduced cell proliferation in a dose-dependent manner (supplementary material Fig. S3B,C). To evaluatethe efficacy of this compound with respect to in vivo metastasis,orthotopic implantation of the highly metastatic 4T1 cell line into themammary fat pad of female BALB/cJ mice was performed aspreviously described (Lee et al., 2014). Following orthotopicimplantation of the cells, the mice were treated with dailyintraperitoneal injections of saline or 1 mg/kg doses of SU6656 for5 days. At 4 weeks post-implantation, tumor burdens were measuredand pulmonarymetastaseswere quantified.Mice treatedwith SU6656displayed significant decreases in surface metastases compared withthe saline-injected mice (Fig. 7A). This result not only confirmed therole of Src activity in tumor metastasis, but also the efficacy of Srcinhibitor SU6656 to preventmetastases in vivo. This decrease in tumormetastasis was associated with a modest increase in primary tumorburden based on the averageweight of the tumors (Fig. 7B). There aremany possible explanations for this observation, including thatSU6656 inhibits cells migrating to form metastases but does nothave a major impact on overall cell viability. This experimentdemonstrates that compounds or gene targets identified by inhibiting

Fig. 5. Disruption of src and tks5a genes by CRISPR/Cas9affects the PLLp migration. Two sgRNAs targeting two differentexons (25 ng/µl) within each gene were co-injected with Cas9mRNA (300 ng/µl) into cldnb:EGFP embryos and the embryosraised. Injected fish were inbred and at 48 hpf WT embryos showthe normal pattern of the migrating primordium and neuromastsdeposition (A), whereas src CRISPR (B), tks5a CRISPR (C) andtks5a−/− (D) embryos show a disruption of the primordiummigratory ability and PLL morphogenesis. Arrows indicate theposition of the primordium in tks5a−/− embryos; asterisk indicatesthe position of the last deposited neuromast. (E) Sequenceconfirmation of tks5a−/−alleles. Allele 1 has an 8-bp insertion and a1-bp deletion, which changed amino acid valine at position 171 tothreonine and truncates the protein after 27 amino acids. Allele 2has a 21-bp insertion and a 4-bp deletion, which changed aminoacid proline at position 369 to glutamine and truncated the proteinafter 14 amino acids.

571

RESEARCH ARTICLE Disease Models & Mechanisms (2015) 8, 565-576 doi:10.1242/dmm.018689

Disea

seModels&Mechan

isms

migration of the zebrafish PLLp can also have strong anti-metastaticactivity in mammals in vivo.

DISCUSSIONIn the last 10 years, zebrafish has emerged as a leadingmodel for drugdiscovery based on phenotype. New transgenes and powerfulimaging technologies have increased the sensitivity and throughputto a degree that nowmakes it possible to screen thousands oreven tensof thousands of compounds. Recent phenotype-driven screens haveproven the efficacyof chemical screeningusing zebrafish, identifyingsmall molecules that regulate signaling pathways, development anddisease states (Jung et al., 2012; Sukardi et al., 2011; Taylor et al.,2010; Terriente and Pujades, 2013; Walker et al., 2012), whereashigher throughput, target-based screening on thewhole has proven tobe a disappointment (Swinney and Anthony, 2011).To identify inhibitors of cell migration and, ultimately, anti-

metastatic compounds, we developed a high-throughput-screeningstrategy combined with automated fluorescence microscopy to

visualize the migration of the posterior lateral primordium. We tookadvantage of the cldnb:EGFP transgenic line, and tested theLOPAC1280, PKIS and NatProd collections to identify compoundsthat inhibited collective cell migration.

A total of 5.57% (n=165) of the compounds that we testedinhibited the formation of the posterior lateral line at concentrationsfrom 0.5-10 μM, without being overtly toxic to the fish. However, itis possible that additional compounds would have inhibitedmigration if tested at concentrations higher than 10 μM, or mighthave inhibited motility at lower (sub-lethal) concentrations. Despitethese caveats, the fact that almost 80% of the compounds tested werenot toxic, and that a manageable number of compounds inhibitedPLL formation suggests that 0.5, 1, 5 and 10 μMwas an appropriaterange of concentrations used in the screen.

As expected, a number of compounds previously described tocontribute to tissue invasion and matrix remodeling were detected inour experiment (Chua et al., 2012; Quintavalle et al., 2011). In vitrohigh-throughput screening showed that inhibitors of Src (SU6656,

Fig. 6. Overexpression of srcmRNA rescues PLLp migration disrupted by Src signaling pathway inhibitors. Phenotypes of 48 hpf cldnb:EGFP embryosupon overexpression of src followed by treatment with Src inhibitors that had been identified in the screen. Overexpression of full-length (fl) src in embryosfollowing DMSO treatment show a normal lateral line development (G). Overexpression of src partially rescued primordium migration defects in embryos treatedwith RBI (D) and SU6656 (F) compared with non-injected-embryos (C,E), respectively. This phenotype was not rescued in Emodin-treated embryos uponoverexpression of src (A,B). (H) Quantification of the average primordium position in drug-treated embryos upon src overexpression relative to the distance fromthe otic vesicle (ov) to the tip of the tail. Experiments were carried out with ten larvae per condition. Error bars indicate +s.d. ***P<0.001; *P<0.05. mRNA(100 pg/2 nl) was injected into one-cell stage embryos. n.i., non-injected-embryo.

572

RESEARCH ARTICLE Disease Models & Mechanisms (2015) 8, 565-576 doi:10.1242/dmm.018689

Disea

seModels&Mechan

isms

RBI), MEK (UO126) and CDK (purvalanol A, GCP74514A,kenpaullone) all inhibited PLLp migration and, possibly, inhibitedthe formation of podosomes and/or invadopodia (Quintavalle et al.,2011).To provide an independent verification of our results, we performed

an in vivo inflammation assay based on neutrophil migration to sitesof injury induced by noninvasive damage to lateral line neuromasts(d’Alencon et al., 2010). All the candidate drugs tested in thisindependent cellmigration assaywere confirmed to possessmigration-inhibitory properties. Even though inflammation is based on therecruitment of individual neutrophils to the site of wounding orinfection– andnot coordinated and cohesive collective cellmigration –in both individual and collective cell migration, cells need adhesion to,and traction on, the substratum, and progress through the extracellularmatrix driven by distant guidance signals. These biophysicalcharacteristics are regulated by similar molecular mechanisms andare induced through multiple signaling pathways.Furthermore, to provide additional evidence for the success of our

model, we performed in vitro migration/invasion assays using thehuman melanoma cell line 501-mel. Many of the tested compoundswere able to inhibit the transwell migration of these cells, as well asMatrigel invasion but, notably, some compounds only appeared toinhibit migration in vivo and would have been missed had they beenscreened for using in an in vitro migration assay (supplementarymaterial Table S3).

Src plays a critical role in a variety of cellular signal transductionpathways, regulating diverse processes such as cell proliferation,motility, adhesion, angiogenesis and survival (Kim et al., 2009).The proto-oncogene c-Src (Src) is a non-receptor tyrosine kinasewhose expression and activity correlates with tumor progression,advanced malignancy and poor prognosis in a variety of humancancers (Chua et al., 2012; Homsi et al., 2007). Although theSrc pathway is certainly already a heavily studied one in cancerresearch, it provided an excellent opportunity to demonstrate thatzebrafish PLLp migration can be used to effectively identify anti-metastatic compounds that work in vivo. Three of the compoundsdetected in our screen are inhibitors of the Src pathway: Emodin,SU6656 and 7-cyclopentyl-5-(4-phenoxy)phenyl-7H-pyrrolo[2,3-d]pyrimidin-4-ylamine (RBI). In all cases, we observed a disruptionin the migration of the PLLp. None of these compounds had beenpreviously reported as an inhibitor of PLLp migration, nor had srcactivity been implicated in PLLpmigration. However, recent studieshave shown that the Src substrate and adaptor protein Tks5 plays arole in the migration of the zebrafish neural crest cells by generatingactin-rich pro-migratory structures (Murphy et al., 2011) and isimplicated in regulating invasive behavior of macrophages (Burgeret al., 2011), suggesting that blocking Src has a general impact oncell migration. Moreover, the Src-Tks5 pathway regulates theformation of podosomes and invadopodia, degradation of the ECM,and the invasion of cancer cells in vivo and in vitro (Blouw et al.,2008; Courtneidge, 2012). Our data demonstrate that the Src-Tks5pathway is directly involved in PLLp migration. Knockout of srcand tks5 through CRISPR/Cas9 targeting resulted in a reduction ofprimordium size and loss of migration. Moreover, overexpression ofsrc rescued the lateral line phenotypes induced by treatment withRBI or SU6656. Thus, taken together these results demonstrate thatthe Src pathway is the true target for these inhibitors and Src plays arole in PLLp migration.

We showed that the Src inhibitor SU6656 significantly reducedtumor metastasis in mice, demonstrating that similar mechanismsare used to control both cell migration during PLLp migration inzebrafish and cancer metastasis. These findings, collectively,support other investigations of Src and its binding protein Tks5,both as markers of invasive disease and as potential therapeutictargets (Arai et al., 2012; Blouw et al., 2008).

Although the demonstration that Src signaling is necessary forcell migration and metastasis may not be surprising, it demonstratesthat the screen can effectively identify compounds with the desiredanti-metastatic properties that will work in vivo. Therefore, ourstudy provides a novel set of candidate molecules that block cellmigration, and identifies (often surprising) target genes withpotentially important roles in collective cell migration in bothdevelopment and disease. These genes range from those involvedin NO synthesis or neurotransmitter signaling, to unknown andpotentially novel targets of natural compounds. Using this in vivoscreening assay to screen libraries of pharmacologically activemolecules of known bioactivity, or libraries designed to targetspecific classes of enzyme, such as kinases, allows rapididentification and direct testing of chemical targets in vivo. Thenew targeted inactivation of genes in zebrafish by using CRISPR/Cas9 provides rapid and powerful confirmation of the bona fidetarget of the inhibitor. Using an in vivo model such as zebrafishinstead of cell-culture systems will also allow for a more rapidand successful transition to pre-clinical mammalian models and,ultimately, to new chemotherapeutic treatments in humans.

In conclusion, we have shown that PLLpmigration can be used asan effective model to rapidly identify potent compounds that

Fig. 7. Inhibition of Src activity by SU6656 decreases tumor metastasisin vivo. After orthotopic implantation of metastatic cells into the mammary fatpad of BALB/cJ mice and 5 days of daily intraperitoneal injections of salinesolution or 1 mg/kg SU6656, tumor burden at the site of injection andpulmonary metastases were assessed (saline, n=13; 1 mg/kg, n=15).Quantification as individual points for pulmonary surface metastases (A) andprimary tumor burden (B) are shown. Both measurable differences arestatistically significant, as indicated by P-values.

573

RESEARCH ARTICLE Disease Models & Mechanisms (2015) 8, 565-576 doi:10.1242/dmm.018689

Disea

seModels&Mechan

isms

comprise anti-metastatic activity, and the effectiveness of thesecompounds can be demonstrated in mammalian tumor models.

MATERIALS AND METHODSZebrafish husbandryAll animal experiments adhered to the NIH Guide for the Care and Use ofLaboratory Animals. Zebrafish embryos and larvae of the cldnb:EGFP (Haasand Gilmour, 2006) and Tg(mpx:GFP) (Renshaw et al., 2006) strains weremaintained in our facility according to standard procedures (Westerfield,2000). All embryos were collected by natural spawning, staged according toKimmel et al. (1995) and raised at 28.5°C in E3 medium (5 mM NaCl,0.17 mMKCl, 0.33 mMCaCl2, 0.33 mMMgSO4 and 0.1%Methylene Blue)in Petri dishes, as described previously (Haffter and Nusslein-Volhard, 1996).Embryonic age is expressed in hours post-fertilization (hpf).

Chemical library screeningThe LOPAC1280 library (Sigma-Aldrich), composed of 1280 bioactivecompounds, the NatProd library (MicroSource Discovery Systems Inc.),containing 800 pure natural products and their derivatives, and the GSKPublished Kinase Inhibitor Set (PKIS), containing 880 compounds, wereselected for the cldnb:EGFP embryo-based screenings. A total of 2960compoundswere screened. 200-nl aliquotsof the chemicals (10 mMinDMSO)were transferred by ultrasonic deposition using anATS-110 (EDCBiosystems)from the mother plates into 96-well culture plates (Costar #3720) to generatediluted stock plates (200 µl final volume). Next, after manual removal of thechorions, cldnb:EGFP embryos were arrayed manually into the 96-well plates(two embryos per well) in 200 μl of system water at 20 hpf, resulting in a finalscreening concentration of 10 µM. The positive control for this assaywas the c-Met inhibitor K252a (Calbiochem #420298; 0.01 to 10 µM), and the negativecontrol was 1% DMSO. Embryos were treated with the compounds at roomtemperature over 2 days. At 36 and 48 hpf, embryos were anesthetized with0.016% Tricaine, and a fluorescence image of individual wells was takenautomatically (magnification 4×) using an iCys research imaging cytometer(MolecularDevices). Imageswere captured andcompiled forvisual inspection.Drugs were considered to be active when the two embryos displayed the samephenotype (edema, lethality orother phenotypes); all putative hits were retestedand comparable results were obtained in each case.

Secondary chemical screeningThe compounds that induced developmental toxicity effects and causedembryo death in the primary screening were re-tested in cldnb:EGFPembryos by using final screening concentrations of 0.5, 1 and 5 μM.Screening procedures were carried out as described above.

Confirmatory testsIn follow-up studies, 20 compounds from the LOPAC1280 and the NatProdlibraries that induced lateral line phenotypes were tested in cldnb:EGFPembryos. Chemical treatment of embryos (n=15 per well) was initiated at 20or 36 hpf and finished at 48 hpf. Embryos were then anesthetized with0.016% Tricaine and fluorescence images of embryos were taken using aninverted Zeiss AXIOVERT200M microscope equipped with an Apotomegrid confocal system.

Chemically induced inflammation assaysChemically induced inflammation (ChIn) assays were performed asdescribed in d’Alencon et al. (2010). Briefly, 15 transgenic Tg(mpx:GFP)larvae at 72 hpf were transferred to 12-well plates in a volume of 1 ml of E3solution. For inhibition, test reagents were added to reach the requiredconcentration to the well. Positive and negative control larvae wereincubated with 1% of DMSO. Incubation with DMSO control and testeddrugs was carried out for 30 min prior to addition of 10 μM CuSO4; thelatter was added directly to the wells that contained the experimental andcontrol larvae for 40 min at 28°C. Larvae were then fixed with 4%paraformaldehyde prepared in phosphate-buffered saline (PBS) andincubated for 1 h at room temperature. After fixation, larvae were washedthree times for 5 min each in PBS-Tween20 with gentle agitation. Countingof leukocytes was done using an inverted Zeiss fluorescence microscope.

Cell culture501-mel human melanoma cells were cultured in DMEM (Invitrogen)supplemented with 10% fetal bovine serum (Invitrogen), antibiotic-antimycotics (100 U/ml penicillin, 100 µg/ml streptomycin, 250 ng/mlamphotericin; Invitrogen) and 2 mM L-glutamine (Invitrogen).

Transwell migration assayCell motility was determined in vitro using a Transwell chamber (BDBiosciences) according to the manufacturer’s instructions. Briefly, 8-mmpore cell-culture inserts were rehydrated for 2 h in 37°C serum-free DMEMprior to use. Then 2.5×105 501-mel human melanoma cells were placed onthe upper chamber in 500 μl serum-free DMEM plus 0.1% DMSO controlor a test drug. In the lower chamber, 750 μl DMEM plus 10% FBS wasadded. After 16 h of incubation at 37°C in 5% CO2, cells from the uppersurface of the membranes were removed by gentle swabbing, and the cellsthat had migrated through the pores were fixed, stained and counted.Relative migration was based on the average number of cells on theunderside of the membrane in five random images generated at 200×magnification per chamber of two independent experiments, and wasnormalized to the results from DMSO control cells.

Cell invasion assayInvasion of cells intoMatrigel was determined in vitro using 24-well BioCoatMatrigel inserts (Becton Dickinson) as described in McGill et al. (2006).Briefly, the invasion chambers were prehydrated with serum-free DMEM(500 μl/well) for 2 h of incubation at 37°C in 5% CO2. After trypsinization,501-mel melanoma cells (5×105) were suspended in 500 μl serum-freemedium and incubated with 0.1% DMSO control or different inhibitors for20 min, and then placed in the upper compartment of the plates. Subsequently,the lower compartment was filled with 10% FBS medium (750 μl) includingrecombinant human HGF (50 ng/ml). After 24 h of incubation, cells werefixed, stained and counted. Non-migratory cells on the upper filter surfacewere removed using a cotton swab, and the total number of invasive cells wascounted at 200× magnification using a phase-contrast microscope. Relativeinvasion was based on the average number of cells on the underside of themembrane in five random images of two independent experiments, and wasnormalized to the results from DMSO-control cells.

Targeting of src and tks5a genes with the CRISPR/Cas9 systemWe generated src and tks5a knockouts using the CRISPR/Cas9 targetingsystem, by targeting two different exons in each gene. The targetingsequences were as follows: src target 1: 5′-GGATTTCCTGAAAGG-TGACA-3′, src target 2: 5′-GGCACCGTCTGACTCCATCC-3′, tks5atarget 1: 5′-GGTGCTGGAGCAGTACGTGG-3′, tks5a target 2: 5′-GGAG-ATGTGGCGCTCAGGGG-3′.

Single-guideRNAs (sgRNAs)were synthesized byannealing and extendingtwo oligonucleotides, and mRNA was transcribed from assembledoligonucleotides using the T7 RNA synthesis kit (Varshney et al., 2015).The cas9 mRNA was prepared by in vitro transcription from the pT3TS-nls-zCas9-nls plasmid (Jao et al., 2013). We co-injected two sgRNAs (25 ng/µl)and 300 ng/µl Cas9 in one-cell stage cldnb:EGFP embryos. Injected embryoswere raised to sexual maturity. The founder fish were out-crossed with wild-type fish, and the germline mutations were identified using fluorescence PCR(Sood et al., 2013) and sequencing. Phenotypes were also scored in someembryos 48 h after injection of CRISPR/Cas9-targeting RNAs.

Microinjection of synthetic src mRNA, rescue experiments100 pg of mRNA in water was injected into one-cell stage cldnb:EGFPembryos. We used full-length src cDNA (Clone ID: 8104558, Dharmacon)and capped synthetic mRNAs were prepared using the T7 mMessagemMachine kit (Ambion). For rescue experiments, chemical treatment ofsrc-mRNA-injected embryos (n=10 per well) with the Src inhibitorsEmodin, RBI or SU6656, or with a DMSO control, was initiated at 20 hpfand finished at 2 dpf. Embryos were then anesthetized with 0.016%Tricaine, and fluorescence images of embryos were captured using aninverted Zeiss AXIOVERT200M microscope and quantified for PLLpmigration.

574

RESEARCH ARTICLE Disease Models & Mechanisms (2015) 8, 565-576 doi:10.1242/dmm.018689

Disea

seModels&Mechan

isms

Orthotopic mammary fat pad injections and SU6656 treatmentFemale BALB/cJ mice were purchased from Jackson Laboratory (BarHarbor, ME). Injections of the mouse mammary tumor 4T1 cells wereperformed as previously described (Lee et al., 2014). Briefly, 105 cells in100 μl saline solution were orthotopically implanted into the mammary fatpads of 10- to 12-week-old female BALB/cJ mice. Beginning on the sameday as the implantations, the mice received daily intraperitoneal injections ofeither saline solution (n=13) or 1 mg/kg SU6656 (n=15) for 5 days. SU6656dosage was determined based on previous publications (Rehni and Singh,2011; Rehni et al., 2012). Three weeks after the implantations, mice wereeuthanized. Tumors were dissected and weighed. Lungs were isolated andsurface metastases enumerated by eye using a dissecting microscope, aspreviously described (Crawford et al., 2007). All animal experiments wereperformed in compliance with the National Human Genome ResearchInstitute Animal Care and Use Committee’s guidelines.

Statistical analysesResults from PLLp migration assays, in vivo ChIns, in vitro migration andinvasion cell assays, and in vivo mouse cancer invasion assays arerepresented as mean±s.d. of independently performed assays. Statisticalanalysis was performed by comparing means of biological replicates usingthe unpaired two-tailed Student’s t-test (Excel, Microsoft). A value ofP<0.05 was considered as significant.

AcknowledgementsWe thank Colin Huck for superior animal care, Julia Cronin for help with cell migrationand invasion assays, and Miguel Allende for help with ChIn assays. Melanoma celllines were kindly provided by Bill Pavan and Yardena Samuels.

Competing interestsThe authors declare no competing or financial interests.

Author contributionsV.E.G. designed experiments, performed all zebrafish experiments and wrote themanuscript. G.K.V. designed and performed zebrafish experiments. M.L. designedand performed mouse experiments, and wrote sections related to mouseexperiments. S.B. performed mouse experiments. L.X. performed zebrafishexperiments. P.S. assisted in screen design and arrayed chemicals for screening.N.P.C. helped to design and discuss mouse experiments. J.I. helped to design drugscreens, provided array assistance and discussed results. S.M.B. helped to designexperiments, discussed data and assisted in writing the manuscript.

FundingThis research was supported by the Intramural Research Program of the NationalHuman Genome Research Institute, National Institutes of Health (S.B., N.P.C.).

Supplementary materialSupplementary material available online athttp://dmm.biologists.org/lookup/suppl/doi:10.1242/dmm.018689/-/DC1

ReferencesAman, A. and Piotrowski, T. (2010). Cell migration during morphogenesis. Dev.Biol. 341, 20-33.

Aman, A. and Piotrowski, T. (2011). Cell-cell signaling interactions coordinatemultiple cell behaviors that drive morphogenesis of the lateral line. Cell Adh. Migr.5, 499-508.

Arai, R., Tsuda, M., Watanabe, T., Ose, T., Obuse, C., Maenaka, K., Minami, A.and Ohba, Y. (2012). Simultaneous inhibition of Src and Aurora kinases bySU6656 induces therapeutic synergy in human synovial sarcoma growth, invasionand angiogenesis in vivo. Eur. J. Cancer 48, 2417-2430.

Blouw, B., Seals, D. F., Pass, I., Diaz, B. and Courtneidge, S. A. (2008). A role forthe podosome/invadopodia scaffold protein Tks5 in tumor growth in vivo.Eur. J. Cell Biol. 87, 555-567.

Burger, K. L., Davis, A. L., Isom, S., Mishra, N. and Seals, D. F. (2011). Thepodosome marker protein Tks5 regulates macrophage invasive behavior.Cytoskeleton 68, 694-711.

Chaffer, C. L. andWeinberg, R. A. (2011). A perspective on cancer cell metastasis.Science 331, 1559-1564.

Chua, K.-N., Sim,W.-J., Racine, V., Lee, S.-Y., Goh, B. C. and Thiery, J. P. (2012).A cell-based small molecule screening method for identifying inhibitors ofepithelial-mesenchymal transition in carcinoma. PLoS ONE 7, e33183.

Courtneidge, S. A. (2012). Cell migration and invasion in human disease: the Tksadaptor proteins. Biochem. Soc. Trans. 40, 129-132.

Crawford, N. P. S., Qian, X., Ziogas, A., Papageorge, A. G., Boersma, B. J.,Walker, R. C., Lukes, L., Rowe,W. L., Zhang, J., Ambs, S. et al. (2007). Rrp1b, anew candidate susceptibility gene for breast cancer progression and metastasis.PLoS Genet. 3, e214.

d’Alencon, C. A., Pena, O. A., Wittmann, C., Gallardo, V. E., Jones, R. A., Loosli,F., Liebel, U., Grabher, C. and Allende, M. L. (2010). A high-throughputchemically induced inflammation assay in zebrafish. BMC Biol. 8, 151.

Friedl, P. and Gilmour, D. (2009). Collective cell migration in morphogenesis,regeneration and cancer. Nat. Rev. Mol. Cell Biol. 10, 445-457.

Gallardo, V. E., Liang, J., Behra, M., Elkahloun, A., Villablanca, E. J., Russo, V.,Allende, M. L. and Burgess, S. M. (2010). Molecular dissection of the migratingposterior lateral line primordium during early development in zebrafish. BMC Dev.Biol. 10, 120.

Ghysen, A. and Dambly-Chaudiere, C. (2007). The lateral line microcosmos.Genes Dev. 21, 2118-2130.

Grant, K. A., Raible, D. W. and Piotrowski, T. (2005). Regulation of latent sensoryhair cell precursors by glia in the zebrafish lateral line. Neuron 45, 69-80.

Haas, P. and Gilmour, D. (2006). Chemokine signaling mediates self-organizingtissue migration in the zebrafish lateral line. Dev. Cell 10, 673-680.

Haffter, P. and Nusslein-Volhard, C. (1996). Large scale genetics in a smallvertebrate, the zebrafish. Int. J. Dev. Biol. 40, 221-227.

Homsi, J., Cubitt, C. and Daud, A. (2007). The Src signaling pathway: a potentialtarget in melanoma and other malignancies.Expert Opin. Ther. Targets 11, 91-100.

Hwang, W. Y., Fu, Y., Reyon, D., Maeder, M. L., Kaini, P., Sander, J. D., Joung,J. K., Peterson, R. T. and Yeh, J.-R. J. (2013). Heritable and precise zebrafishgenome editing using a CRISPR-Cas system. PLoS ONE 8, e68708.

Jao, L.-E., Wente, S. R. and Chen, W. (2013). Efficient multiplex biallelic zebrafishgenome editing using a CRISPR nuclease system. Proc. Natl. Acad. Sci. USA110, 13904-13909.

Jung, D.-W., Oh, E.-S., Park, S.-H., Chang, Y.-T., Kim, C.-H., Choi, S.-Y. andWilliams, D. R. (2012). A novel zebrafish human tumor xenograft model validatedfor anti-cancer drug screening. Mol. Biosyst. 8, 1930-1939.

Kim, L. C., Song, L. and Haura, E. B. (2009). Src kinases as therapeutic targets forcancer. Nat. Rev. Clin. Oncol. 6, 587-595.

Kimmel,C.B.,Ballard,W.W.,Kimmel,S.R.,Ullmann,B.andSchilling,T. F. (1995).Stages of embryonic development of the zebrafish. Dev. Dyn. 203, 253-310.

Lee, M., Dworkin, A. M., Gildea, D., Trivedi, N. S., Moorhead, G. B. andCrawford, N. P. S. (2014). RRP1B is a metastasis modifier that regulates theexpression of alternative mRNA isoforms through interactions with SRSF1.Oncogene 33, 1818-1827.

Matsuda, M., Nogare, D. D., Somers, K., Martin, K., Wang, C. and Chitnis, A. B.(2013). Lef1 regulates Dusp6 to influence neuromast formation and spacing in thezebrafish posterior lateral line primordium. Development 140, 2387-2397.

McGill, G. G., Haq, R., Nishimura, E. K. and Fisher, D. E. (2006). c-Met expressionis regulated by Mitf in the melanocyte lineage. J. Biol. Chem. 281, 10365-10373.

Murphy, D. A., Diaz, B., Bromann, P. A., Tsai, J. H., Kawakami, Y., Maurer, J.,Stewart, R. A., Izpisua-Belmonte, J. C. and Courtneidge, S. A. (2011). A Src-Tks5 pathway is required for neural crest cell migration during embryonicdevelopment. PLoS ONE 6, e22499.

Parmar, V. S., Jain, R., Sharma, S. K., Vardhan, A., Jha, A., Taneja, P., Singh, S.,Vyncke, B. M., Bracke, M. E. and Mareel, M. M. (1994). Anti-invasive activity of3,7-dimethoxyflavone in vitro. J. Pharm. Sci. 83, 1217-1221.

Quintavalle, M., Elia, L., Price, J. H., Heynen-Genel, S. and Courtneidge, S. A.(2011). A cell-based high-content screening assay reveals activators andinhibitors of cancer cell invasion. Sci. Signal. 4, ra49.

Rehni, A. K. and Singh, N. (2011). Modulation of src-kinase attenuates naloxone-precipitated opioid withdrawal syndrome in mice. Behav. Pharmacol. 22, 182-190.

Rehni, A. K., Singh, T. G. and Arora, S. (2012). SU-6656, a selective Src kinaseinhibitor, attenuates mecamylamine-precipitated nicotine withdrawal syndrome inmice. Nicotine Tob. Res. 14, 407-414.

Renshaw, S. A., Loynes, C. A., Trushell, D. M. I., Elworthy, S., Ingham, P.W. andWhyte, M. K. B. (2006). A transgenic zebrafish model of neutrophilicinflammation. Blood 108, 3976-3978.

Shrimali, D., Shanmugam, M. K., Kumar, A. P., Zhang, J., Tan, B. K. H., Ahn,K. S. and Sethi, G. (2013). Targeted abrogation of diverse signal transductioncascades by emodin for the treatment of inflammatory disorders and cancer.Cancer Lett. 341, 139-149.

Sood, R., Carrington, B., Bishop, K., Jones, M., Rissone, A., Candotti, F.,Chandrasekharappa, S. C. and Liu, P. (2013). Efficient methods for targetedmutagenesis in zebrafish using zinc-finger nucleases: data from targeting of ninegenes using CompoZr or CoDA ZFNs. PLoS ONE 8, e57239.

Sukardi, H., Chng, H. T., Chan, E. C. Y., Gong, Z. and Lam, S. H. (2011). Zebrafishfor drug toxicity screening: bridging the in vitro cell-based models and in vivomammalian models. Expert Opin. Drug Metab. Toxicol. 7, 579-589.

Swinney, D. C. and Anthony, J. (2011). Howwere newmedicines discovered?Nat.Rev. Drug Discov. 10, 507-519.

Taylor, K. L., Grant, N. J., Temperley, N. D. and Patton, E. E. (2010). Smallmolecule screening in zebrafish: an in vivo approach to identifying new chemicaltools and drug leads. Cell Commun. Signal. 8, 11.

575

RESEARCH ARTICLE Disease Models & Mechanisms (2015) 8, 565-576 doi:10.1242/dmm.018689

Disea

seModels&Mechan

isms

Terriente, J. and Pujades, C. (2013). Use of zebrafish embryos for small molecule

screening related to cancer. Dev. Dyn. 242, 97-107.Valastyan, S. and Weinberg, R. A. (2011). Tumor metastasis: molecular insights

and evolving paradigms. Cell 147, 275-292.Varshney, G. K., Pei, W., LaFave, M. C., Idol, J., Xu, L., Gallardo, V., Carrington,

B., Bishop, K., Jones, M., Li, M. et al. (2015). High-Throughput Gene Targeting

and Phenotyping in Zebrafish Using CRISPR/Cas9. Genome Res. (in press).Walker, S. L., Ariga, J., Mathias, J. R., Coothankandaswamy, V., Xie, X., Distel,

M., Koster, R. W., Parsons, M. J., Bhalla, K. N., Saxena, M. T. et al. (2012).

Automated reporter quantification in vivo: high-throughput screening method forreporter-based assays in zebrafish. PLoS ONE 7, e29916.

Wei, W. T., Lin, S. Z., Liu, D. L. andWang, Z. H. (2013). The distinct mechanisms ofthe antitumor activity of emodin in different types of cancer (Review). Oncol. Rep.30, 2555-2562.

Westerfield, M. (2000). The Zebrafish Book: A Guide for the Laboratory Use ofZebrafish (Brachydanio rerio), 4th edn (ed. M. Westerfield). Eugene, OR:Distributed by the Institute of Neuroscience, University of Oregon, Chapters 1-3.

Yang, J. and Weinberg, R. A. (2008). Epithelial-mesenchymal transition: at thecrossroads of development and tumor metastasis. Dev. Cell 14, 818-829.

576

RESEARCH ARTICLE Disease Models & Mechanisms (2015) 8, 565-576 doi:10.1242/dmm.018689

Disea

seModels&Mechan

isms